varprox 1.0.6

Optimization algorithms for solving penalized non-linear least squares problems

The Package varprox is designed for solving penalized separable non-linear least squares problems. It extends the standard variable projection method by adding regularization on the non-linear variable. More details can be found in the documentation.

Package features

• Methods for minimizing separable non-linear least squares problems with penalizations and box constraints on variables.

• Non linear model fitting in engineering.

• Applications to the statistical inference for stochastic processes.

Installation from sources

The package source can be downloaded from the repository.

The package can be installed through PYPI with

pip install varprox

Communication to the author

varprox is developed and maintained by Arthur Marmin and Frédéric Richard. For feed-back, contributions, bug reports, contact directly the author, or use the discussion facility.

Licence

varprox is under licence GNU GPL, version 3.

Citation

When using varprox, please cite the papers

1. Escande, P. and Richard, F. Full inference for the anisotropic fractional Brownian field. Journal of Probability Theory and Mathematical Statistics, 110:13–29, 2024. doi:10.1090/tpms/1204.

2. Marmin, A. and Richard, F. Varprox: a primal-dual variable projection method for the minimization of penalized separable non-linear least squares. Inverse Problems, 2025. doi:10.1088/1361-6420/ae0e48.

Credits

Varprox is written and maintained by Arthur Marmin and Frederic Richard, Aix-Marseille University, France.